Os calcis fractures: Analysis of interobserver variability in using sanders classification☆

Article Outline

• Abstract
• Materials and methods
• Results
• Discussion
• References
• Copyright

Abstract 

The os calcis is the most frequently fractured tarsal bone. In 1992 Sanders developed a classification system based on coronal and axial computed tomography (CT) scans of the calcaneus. This classification is the one used most frequently today in treatment decision making and reporting of results. The objective of this study was to assess the degree of interobserver variability in using this classification system. Thirty CTs of calcaneal fractures were chosen randomly from the past 5 years in 2 tertiary care centers. The CTs were reviewed by 3 orthopedic surgeons and one senior orthopedic resident who classified the fractures according to Sanders' classification. The results were first tabulated and analyzed by using a weighted κ test including the subcategories. The weighted κ value achieved was .56, with a 95% confidence interval of .45–.67. The subcategories of the classification were then further combined and a second weighted κ test was performed to assess agreement between general classes. The weighted κ value achieved was .48, with a 95% confidence interval of 0.37–0.59. We concluded that Sanders' classification system did prove to achieve moderate agreement among users, thus representing a useful classification system. (The Journal of Foot & Ankle Surgery 42(1):21–23, 2003)

Keywords:  os calcis fractures, calcaneus fractures, Sanders classification, CT scan

The os calcis is the most frequently fractured tarsal bone ⁽¹⁾. The complications resulting from such a fracture are numerous and include malunion, posttraumatic subtalar arthritis, chronic foot pain, peroneal tendonitis, and lateral impingement syndrome ⁽²⁾. There have been many attempts at accurately describing and classifying fractures of the os calcis. Classifications range from using mechanism of injury as Bohler did in the 1930s to Essex-Lopresti's analysis of location of the fracture ³, ⁴. Despite exhaustive methods of analyzing and classifying these fracture patterns there were often less than ideal outcomes for patients with intra-articular fractures of the os calcis. Surgeons were often in disagreement as to which classification to use as well as when and which approach to use in surgical intervention of these fractures ⁽³⁾.

In an attempt to reduce confusion and clarify classifications of these fractures, Sanders developed a classification system of intra-articular fractures based on coronal and axial computed tomography (CT) scans of the fracture ⁽³⁾. This scheme is based on coronal and axial CT scans of the calcaneus (Fig. 1).

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 1. 

  Sanders classification of intra-articular fractures of the os calcis based on CT scans. Reprinted with permission ⁽⁵⁾.

Sanders et al. described using the CT cut that displayed the undersurface of the posterior facet of the talus, which was the widest ⁽³⁾. Subsequently, the posterior facet was divided into 3 sagittal columns of equal size by 2 lines (A and B), which divide the posterior facet into 3 columns, namely: medial, lateral, and central ⁽³⁾. A third line, C, also is considered. It runs along the medial edge of the posterior facet and represents a line separating the sustenaculum tali from the medial edge of the posterior facet ⁽³⁾. Consequently, 3 potential primary fracture lines are created, beginning laterally with A and moving medially to C, thereby creating 4 potential fracture fragments overall.

The classification is divided into 4 types. Type I represents all undisplaced fractures, regardless of the number of pieces involved. Type II represents 2-part fractures of the posterior facet and can be divided into 3 subtypes based on the primary fracture line, and include IIA, IIB, and IIC (Fig. 1) ⁽³⁾. Type III represents 3-part fractures with a centrally depressed fragment. Again, this class can be subdivided into 3 subclasses based on the fracture lines, IIIAB, IIIAC, and IIIBC (Fig. 1). Type IV fractures represent highly comminuted fractures patterns.

This classification was developed in part to help stratify patients and thereby assist in deciding which patients required surgical intervention and which patients could be treated conservatively. Sanders used his classification to help predict the prognosis of these fractures and found that those patients who were best treated surgically were those who had types II and III fracture patterns ⁽⁵⁾. Given that Sanders's classification is used frequently by surgeons and the classification may indeed effect potential therapeutic decisions, it is essential that the classification system be applied uniformly and consistently. Hence the purpose of this study was to assess the degree of interobserver variability in using Sanders's classification system.

Back to Article Outline

Materials and methods 

Fifty CT scans of calcaneal fractures were reviewed from the databases between 1995 and 2000. Twenty-nine patients representing 30 intra-articular calcaneal fractures were selected on the basis that they represented fractures that were classifiable according to Sanders's classification. The CT scans were selected by the author who was not a reviewer and thus independent of the results. Although the CT scans were selected from the institution where the reviewing surgeons practice, the reviewers were blinded as to patients' names, treating surgeons, and the treatments the patients received.

The CT scans, including the entire sheet of slices, were distributed to 3 orthopedic surgeons from the Department of Orthopaedic Surgery at our institution, as well as 1 senior orthopedic resident. The participants also were provided with a figure describing Sanders's classification and asked to classify each fracture based on Sanders's classification.

The results were tabulated subsequently excluding 1 of the CTs after it was realized that it was in fact a sustenaculum fracture and not classifiable with the index classification. The classes and subclasses were treated as ordinal values and, consequently, a weighted κ test was chosen as the analysis of choice. Two separate weighted κ tests were performed using first the entire class and then the subclasses. Ninety-five percent confidence intervals were calculated for both weighted κ values.

A weighted κ value is a test of partial agreement between observers. It is most useful in categoric data. The amount of agreement is assigned a specific weight and subsequently a κ value is calculated. Values range from −1 to +1, with a value of 0 representing the agreement by chance alone ⁽⁶⁾. When one assumes that competent observers are used, a weighted κ value approaching +0.5 to 0.6 represents an acceptable degree of agreement ⁽⁶⁾.

Back to Article Outline

Results 

The 4 surgeons' results were tabulated in terms of both classes and subclasses and the results are tabulated in Tables 1 and 2.

Table 1. Total number of CT scans classed according to types based on Sanders's classification according to individual observer

Table 2. Total number of CT scans classed according to subtypes based on Sanders's classification according to individual observer

The weighted κ value obtained using the classes as a whole was .48, with a standard error of .058 and a subsequent confidence interval of .37–.59. The weighted κ value obtained maintaining the subtypes was .56, with a standard error of .058 and subsequent confidence interval of .45–.67.

Back to Article Outline

Discussion 

The purpose of any clinical classification system is not only to group and organize but also to allow physicians to communicate with a common understanding of the classes, thereby influencing assessments and potential treatments. Like any classification system, to facilitate discussion of potential treatment options for a given fracture of the os calcis, it is important that interpretation of the classification be uniform and universal. Sanders's classification attempts to achieve this goal and potentially allows surgeons to alter potential surgical interventions based on the classification. It was our hypothesis that the Sanders classification did in fact achieve this goal and that there was reasonable interobserver agreement in using the classification system. The results we obtained supported our hypothesis.

We decided to assess interobserver variability both in terms of the classes themselves in addition to the subclasses. We felt that this was essential to evaluate variability in the overall classes, as well as to evaluate the variability in using subtypes to assess fully the use of Sanders's classification.

The results obtained for the amount of variability between users for the classes as a whole revealed a weighted κ value of .48, which represents a moderate strength of agreement ⁽⁶⁾. Consequently, one can deduce that there was reasonable agreement among surgeons in classifying os calcis fractures irrespective of subtypes.

The results obtained for the amount of variability between users when considering the classes and the subclasses showed a weighted κ value of .56, which represents again a moderate strength of agreement ⁽⁶⁾. Consequently, once again we were able to conclude that there was a reasonable degree of consistency between observers when classifying os calcis fractures by using Sanders's classification in terms of both classes and subclasses.

We believe that the conclusion that the classification system represented consistency between users was heightened by the fact that the users in this study were not all foot and ankle surgeons, or trauma surgeons. The population of surgeons surveyed included a hand and upper limb surgeon, a foot and ankle surgeon, an arthroplasty/sports medicine surgeon, and a senior resident.

The analysis test of choice, the weighted κ test, is reserved for ordinal values. This therefore assumes that a grade III fracture is worse than a grade II fracture. We felt that this was a legitimate assumption and should be brought to the attention of the reader. Although the same assumption does not apply for the subclasses, that is, a grade IIAC is worse than a grade IIAB, a weighted κ test still provides an assessment of the degree of agreement and may in fact underestimate of the degree of agreement. As well, the number of observers chosen was relatively small and introducing more observers may tighten the confidence intervals.

Overall, we concluded that Sanders's classification system proved to achieve moderate agreement among users, thus representing a useful classification system. The level of agreement supports the conclusion that there is consistency and uniformity in the use of the classification.

Back to Article Outline

References 

1. Rockwood CA, Green DP, Bucholz RW. In: 4th ed. Rockwood and Green's Fractures in Adults. vol. 2:Philadelphia: Lippincott-Raven Publishers,; 1998;p. 2325–2326
   • View In Article
2. Myerson M, Quill G. Late complications of fractures of the calcaneus. J Bone Joint Surg. 1993;75A:331–341
   • View In Article
3. Sanders R, Fortin P, DiPasquale T, Walling A. Operative treatment in 120 displaced intra-articular calcaneal fractures; results using a prognostic computed tomography scan classification. Clin Orthop. 1993;290:87–95
   • View In Article
4. Giachino AA, Uhthoff HK. Intra-articular fractures of the calcaneus. J Bone Joint Surg. 1989;71A:784–787
   • View In Article
5. Sanders R. Intra-articular fractures of the calcaneus: present state of the art. J Orthop Trauma. 1992;6:252–265
   • View In Article
   • MEDLINE
   • CrossRef
6. Kramer MS, Feinstein AR. The biostatistics of concordance. Clin Pharmacol Ther. 1981;29:111–123
   • View In Article
   • MEDLINE
   • CrossRef